Keynotes

Hans Kristian Flaatten and Audun Fauchald Strand from Norway treated us to some great insights from their journey in providing a central kubernetes platform for the Norwegian Labor and Welfare Administration, starting back in 2018. After maturing this platform and succeeding in supporting a great number of application teams, they stepped up to deliver this to different branches (aka customers) in the Norwegian government.

Starting and scaling a platform engineering team

Camille Fournier and Ian Nowland gave a wonderful talk about their insights into building, hiring for, and managing platform teams. They had great points about the finding the right balance (having a platform team not too early and not too late), being customer-focussed, how to scope your product, and how to be part of critical operations to ensure budget and HC.

Live migration of a database fleet while serving millions of queries

Joining from Clickhouse, Manish Gill and Jayme Bird took us on a journey to their database migration of thousands of services. They introduced a change to stateful set, using so-called multiSts do have more control about the scaling operations.

The did their migration in a cloud-native way in k8s, using the Clickhouse operator and wrote a custom migration controller.

Naturally, they ran into a suite of issues that required fixes to the operator and controller. The key message here is that even despite extensive testings, this needs to be conducted cluster by cluster, noticing issues and fixing those one at a time. The entire process took 6 around months, and also got them to introduce partial maintenance which restricts cluster scaling, backups etc., but ensure ongoing query capabilities.

Tutorial: Building an AI-powered game

I really enjoyed the Red Hat tutorial on building a ‘simple’ rock-paper-scissors game. Some part of the logic is certainly simple, but the entire stack isn’t, and it was great to get my hands on so many different parts of the tech in such a short time. This covered

• OpenShift, OpenShift AI
• pytorch, onnx, yolo
• kubeflow
• ArgoCD

Combining all of this was great fun, and I look forward to re-using some of this in the future.